Engineering Psychology

1
Engineering Psychology Human Performance

• Outline of Lecture 8
• Review of lecture 7
• Vigilance Mitigations
• Information Theory
• Information Theory determining HSR
• Absolute Judgment and Sorting Tasks
• Displays and Attention

2
Information Theory

• For a group of events, HS and HR are determined
  using the equation for Have
• HT is determined by HT HS HR - HSR
• HSR represents the dispersion of
  stimulus-response relationships (does a
  particular stimulus always elicit the same
  responses?)

3
Information Theory

• HT HS HR - HSR
• Determining HSR (and hence, HT)

HSR
Hloss
HT
noise
HS
HR
4
Information Theory

• Review Computing the quantities of HS, HR

5
Information Theory

• Determining HSR (and hence, HT) compute average
  of values within matrix

HSR log2(4) 2 bits HSR log2(8) 3 bits
6
Information Theory

• Limitations of Information Theory
• HT reflects only the consistency of mappings
  between stimuli and responses not the accuracy or
  appropriateness of the stimulus-response mappings
• Example Abbot and Costello
• Whos on first? vs. Whos on first.
• HT also does not take into account the size of an
  error

7
Perceptual Judgment

• Perceptual Judgment Judgments of stimulus
  magnitudes above threshold
• Contrast with detection stimulus at threshold
• Operator must make a judgment about the magnitude
  of a stimulus that is well above threshold, thus
  detection is not an issue, e.g.,
• Will my car fit in that parking space?
• Do I need to mow the lawn?
• Is it so cold that I need a jacket?

8
Perceptual Judgment

• Unidimensional judgments
• stimuli vary along one dimension only
• observer places stimuli into 2 or more categories
• Multidimensional judgments
• stimuli vary along more than one dimensions
• observer places stimuli into 2 or more categories
  spread across multiple dimensions
• information theory assesses the consistency of
  match between the stimulus and its categorization

9
Unidimensional Judgment

• Channel capacity
• With 5 or more categories errors begin to occur
  much more frequently
• HT lt HS

Maximum capacity
4 categories 2 bits, 8 categories 3 bits
10
Unidimensional Judgment

• Channel capacity (cont.)
• Cause of limited capacity?
• Not sensory
• discrimination performance is typically very good
  for a number of stimulus domains
• difference threshold or just noticeable
  difference (JND) is typically less than 10
• Memory categories must be remembered
• Miller (1956) working memory capacity is
  generally 7 /- 2 items 2-3 bits of information

11
Multidimensional Judgment

• Used when stimuli vary along more than one
  dimension
• Most real-world stimuli are multidimensional
• Independent vs. dependent dimensions
• Independent (orthogonal) change along one
  dimension does not affect the other dimension
• Dependent (correlated) change along one
  dimension is accompanied by change along the
  other dimension

12
Multidimensional Judgment

• Human performance
• higher channel capacity with multidimensional
  information

• Egeth Pachella (1969)
• unidimensional capacity 3.4 bits (10 levels)

• multidimensional capacity 5.8 bits (57 levels)
• dimensions do not sum perfectly, some information
  is lost

13
Multidimensional Judgment

• Additional independent dimensions increase HT but
  with a cost (diminishing returns in bits per
  dimension)

14
Multidimensional Judgment

• Correlated dimensions
• max HS and HT is less due to redundancy, but
• cost (bits/dim.) for extra dimensions is less
• (Note slope representing perfect performance is
  less than that for independent dimensions)

15
Multi-dimensional Judgment and Displays

• Separable vs. integral dimensions
• Separable each dimension can be physically
  specified independent of the other dimension(s)
• Example color and
• fill texture of an object,
• perpendicular vectors

• Integral one dimension must be present for the
  other dimension to be defined
• dimensions are dependent
• Examples color and
• brightness of an object,
• rectangle height width

16
Multi-dimensional Judgment and Displays

• What happens if a display has multiple
  dimensions, but the operator must make a
  unidimensional judgment?
• Real World Examples?
• Laboratory Example Garners sorting task
• Observers sort two-dimensional (or
  multi-dimensional) stimuli into discrete
  categories of a single dimension
• Three conditions
• control
• orthogonal
• redundant

17
Garners Sorting Task Conditions

• control sort along each dimension while ignoring
  the other dimension, which is constant
• dimensions uncorrelated
• e.g., judging height of rectangles of constant
  width and width of rectangles of constant height
• orthogonal sort along each dimension while
  ignoring the other dimension, which varies
• dimensions uncorrelated
• e.g., judging height of rectangles while width
  varies and width of rectangles while height
  varies
• redundantsort along either of two dimensions
• dimensions perfectly correlated
• e.g., judging the width or height of rectangles
  of constant aspect ratio (r 1.0) or area (r
  -1.0)

18
Multi-dimensional Judgment and Displays

• Human Performance for Garners sorting task
• Typical performance
• best redundant sort redundancy gain
• middle control
• worst orthogonal sort orthogonal cost
• Effect of separable or integral dimensions
• integral dimensions (e.g., rectangles) increases
  redundancy gain and orthogonal cost
• separable dimensions (e.g., vectors) minimizes
  redundancy gain and orthogonal cost, BUT...
• gain/cost increased if judged dimension has low
  saliency

19
Multi-dimensional Judgment and Displays

• More on integral dimensions
• integral dimensions can produce emergent
  properties a unidimensional stimulus property
  that results from combining 2 or more dimensions
• redundancy gain or orthogonal cost can depend on
  sign of correlation
• referred to as configural dimensions
• gain/cost depends on saliency of emergent feature
• e.g., rectangles height and width correlation,
  rHW

rHW -1.0, emergent feature shape
rHW 1.0, emergent feature area
20
Multi-dimensional Judgment and Displays Summary

• Stimuli can vary along multiple dimensions
• When operator classifies along all dimensions
• more information can be transmitted
• bits per dimension is less (loss increased)
• greater loss for correlated (dependent)
  dimensions compared to independent dimensions
• When operator classifies along one dimension
• integral displays produce a redundancy gain,
  depending on the emergent properties of their
  configuration
• separable displays can produce a redundancy gain
  if stimuli are difficult to detect (dual coding)

21
Dimensionality and Displays Design Implications

• Industrial Sorting Tasks sorting of products
• often little control over stimulus (the product)
• minimize uncorrelated (irrelevant) dimensions
• Symbolic sorting sorting of information from
  displays
• total control over stimulus (part of the design!)
  
• correlated dimensions represented by integral
  displays can produce emergent features, aiding
  categorization, e.g, temp. pressure in a boiler
• unidimensional judgment impaired by integral
  displays of uncorrelated dimensions

22
Perceptual and Attentional Limitations
23
Perceptual Systems

• How does energy become a perception?
• Perceptual systems act as filters (HT lt HS)
• Some data filtered due to limitations of
  perceptual systems, e.g. resolution of fine
  detail
• bottom-up (or passive) filtering
• Some data filtered due to current goals we
  choose to ignore the data and attend something
  else, e.g., the hardness of the chair youre
  sitting on
• top-down (or active) filtering with attention

24
Filters in the Visual System

• Passive Filters
• Visual Field roughly 150 deg. (varies)
• Useful Visual Field
• Task dependent
• Visual resolution (acuity) is heterogeneous
• Fovea
• Peripheral Vision
• Resolution vs. sensitivity

relative visual acuity
Nasal Blindspot Temporal Degrees
from the fovea
25
Filters in the Visual System

• Passive Filters
• Visual Field roughly 150 deg. (varies)
• Useful Visual Field
• Task dependent
• Visual resolution (acuity) is heterogeneous
• Fovea
• Peripheral Vision
• Resolution vs. sensitivity

26
Filters in the Visual System

• Active Filters
• may be applied voluntarily or involuntarily
• Types
• Movements of the Body
• bring objects into field of view (FOV)
• very slow (relative to head and eye)
• Movements of the Head
• bring objects into FOV and orient fovea toward
  objects of interest
• slow relative to eye movements

27
Filters in the Visual System

• Types of Active Filters (cont.)
• Movements of the Eye - orient fovea more quickly
• Saccades fast (e.g., 100 deg./s), jerky
• 2-4 per second, separated by fixations
• fixation dwell typically related to stimulus
  complexity or difficulty of information
  extraction
• Smooth Pursuit slow (0 -10 deg./s), smooth
• Movements of Focal Attention
• fastest enhancement of processing
• typically lead eye movements

28
Focal Visual Attention

• Function selects (filters) information for
  further processing
• Thought to provide temporal priority or
  additional resources for processing particular
  stimuli
• limited capacity -- unattended stimuli are
  ignored and may not be processed
• difficult to divide attention (especially under
  stress-tunnel vision)

29
Focal Visual Attention

• What is attention?
• Literally?
• ????? (we dont know yet)
• Attention may be the phenomenal experience
  resulting from synchronization of firing of cell
  assemblies in different cortical modules
• Metaphors for understanding attention
• Spotlight -- attention distributed in space
• Zoom lense -- processing advantage trades off
  with spatial extent
• Resources

30
Perceptual and Attentional Limitations
31
Perceptual Filtering and Supervisory Control

• Supervisory Control any task involving scanning
  of displays and selection of relevant stimuli
• Operators display sampling is based on their
  mental model of event probabilities
• High event rate channels sampled more often
• Design to minimize scan times and errors
• locate high event channels centrally
• locate channels with related information (that
  often require sequential sampling) in close
  proximity
• Probability adjustment not as extreme as needed
  for optimal performance (similar to sluggish beta)

32
Perceptual Filtering and Supervisory Control

• Sampling behavior reflects imperfections of human
  memory
• may result in forgetting to sample a channel
• may sample a channel too often
• explains oversampling of low event rate
  channels
• having channels which arent visible (in the
  Norman sense) increases the probability that they
  will be forgotten
• can be overcome with sampling reminders
• Sampling becomes more optimal if preview is
  available -- helps provide accurate mental model

33
To Prepare for Next Class

• If you have not already done so
• Read W3 and W4
• Lecture 9 Topics
• Displays and Attention